How Does My Power Supply Impact Overclocking? @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
How Does My Power Supply Impact Overclocking? - Yesterday I was having a conversation on IRC with a bunch of guys that are very much into building desktop gaming PCs and of course many of these enthusiasts focus on overclocking. I was asked how much a PSU came into importance when we are overclocking our system in terms of overall stability.
 
So... the tl;dr would be, the PSU (as long as it remains within ATX spec) does not affect maximum overclocking, as long as the CPU/GPU VRMs do not overheat. But it does affect long term component lifespan, especially when overclocking.

At least theoretically.
 
But what we'd _really_ like to see is what overclock limits you run into for a given hardware combo and a gold, silver, bronze and fail award PSU. Probably in that order, because the last one might blow stuff up. :D

Obviously this simplifies the longevity issue, but one might expect to see some of the heat issues manifest in a 24 hour stability test.
 
naturbo2000 said:
But what we'd _really_ like to see is what overclock limits you run into for a given hardware combo and a gold, silver, bronze and fail award PSU. Probably in that order, because the last one might blow stuff up. :D

Obviously this simplifies the longevity issue, but one might expect to see some of the heat issues manifest in a 24 hour stability test.
Click to expand...

Watercool dem vrms = overheating vrms solved.
 
Tsumi said:
Watercool dem vrms = overheating vrms solved.
Click to expand...

But if you are using a crappy PSU it is going to make the VRMs work harder and thus produce more heat and not as steady of voltage and more ripple. The VRMs can only regulate up to a certain point.

Past that, it will affect overclocking and even stability at stock settings if the PSU is bad enough.
 
jojo69 said:
AX 1500i


???


where's the review of that?
Click to expand...

Corsair is sending one TODAY after telling us a month ago it would. Don't worry, we purchased one a good while ago to run side by side to see if what is purchased and what is sampled is the same thing.
 
It would be fun to see over clocking results with the most out of spec PSU you've tested, then results with a great PSU.
 
Is there a way to reliably test a PSU for us enthusiasts without dedicated testing tools? I have a 1k PSU that I've been using for what seems like 5 years and this article made me wonder what would happen if it gave out. Hoping it goes quietly in the night and not take out the system... Is there a method to see if age is a factor?

I should point out I have 0 problems so far in a 6month old new build but now I'm paranoid because I push my CPUs.
 
undercoverDrunk said:
Is there a way to reliably test a PSU for us enthusiasts without dedicated testing tools? I have a 1k PSU that I've been using for what seems like 5 years and this article made me wonder what would happen if it gave out. Hoping it goes quietly in the night and not take out the system... Is there a method to see if age is a factor?

I should point out I have 0 problems so far in a 6month old new build but now I'm paranoid because I push my CPUs.
Click to expand...

Things that can be easily caught with common tools would cause computer instability on its own anyways. The only thing you can somewhat measure accurately is voltage with a DMM. Everything else requires an expensive oscilloscope and programmable load tester.
 
So if a motherboard was killing DIMMs would that be a red flag? How could you tell if it was the PSU or just bad memory?

That's the problem with PSU purchases, its like working out your core, you don't think you need it until something breaks.
 
undercoverDrunk said:
So if a motherboard was killing DIMMs would that be a red flag? How could you tell if it was the PSU or just bad memory?

That's the problem with PSU purchases, its like working out your core, you don't think you need it until something breaks.
Click to expand...

That's why you only purchase professionally reviewed power supplies. You don't gamble on unknown power supplies.
 
I double filter the mains power before it reaches the PCs PSU.

If it works for booze it can't hurt my PC!

Always wondered if mains filtering pre PC affects the PSU's power quality delivery.
 
Did you really mean 'viscous circle' and not 'vicious circle'? Because it is somewhat apropos!
 
daglesj said:
I double filter the mains power before it reaches the PCs PSU.

If it works for booze it can't hurt my PC!

Always wondered if mains filtering pre PC affects the PSU's power quality delivery.
Click to expand...

What do you mean double filter? Can you link a "filter"? If it works for booze...heh...
 
Stoly said:
So the final answer is... maybe?
Click to expand...

Pretty much this.In most cases the lifespan of the VRMs will be much more than the useable life of the PC. In most cases the power supplies generally are pretty good these days as long as you didn't find it in a bin in walmart.

What I would like to have seen them do is grab two budget crap PSUs, two mid range PSUs and two PSUs known for really great quality and tight tolerances and then just do some overclocking and see how the max figures turn out and how much change one the VRMs was detected.

Instead we get a "I swear it matters, for sure, you can trust me because I like PSUs".
 
Dayaks said:
What do you mean double filter? Can you link a "filter"? If it works for booze...heh...
Click to expand...

One of these from the mains -

http://www.dem-uk.com/roxburgh/prod...standing_iec_in_line_filter_high_performance/ ( we used to run £10000 video conference systems off them)

Which I then have plugged one of these -

http://tacima.com/mains_cond_cs929.html

When then uses a braided mains cable to the PC with Ferrite cores. I also have ferrite cores on the power leads inside my workstation feeding to the GPU and HDDS.

The workstation has a 875W PSU.

I had the stuff kicking around so why not use it.
 
Last edited:
To customers at work I try to preach that the PSU is the most important overall component in your computer and to not cheap out on it. Doesn't mean you have to buy a $200 PSU for a build with onboard graphics but have enough self respect and smarts to not get a piece of crap.

Just last week I had some moron come in saying his computer wouldn't turn on. It wouldn't turn on because the PSU was dead. The PSU was most likely dead because this genius hooked up his new GTX 750 TI to the 286 watt PSU that came with his prebuilt desktop which had no brand on it I had ever heard of and 11 amps on the +12V rail. It lasted all of a couple weeks before it crapped out on him.
 
Last edited:
It supplies voltage and current.
The extreme states of supply related failure for computers are:
Undersupply – shutdown, Oversupply – damage
Lesser supply related failures are harder to define
Precision – noise, Accuracy – trueness, closeness to spec, the usual multimeter reading
Motherboards and graphics cards have their own set of filters and regulators which help correct these failures. VRMs. CPU, GPU, and RAM don’t use the voltages output directly from the power supply anymore. These regulators have a rather wide tolerance now making the Accuracy of the input of lesser importance. However USB, Ethernet, Audio, and PCIE and other components seem to be unregulated therefore requiring Accuracy.

The real question is: How does noise, or the precision of power, affect overclocking?
By using increased voltage, higher signal strength is generally produced to overcome noise.
But its not just the power supply which produces noise, every single component from fans to wires adds noise. Filters help reduce noise. The best places for filters are near the source, and near the device being powered. Thus by putting filters in the power supply it has a shorter path and is quicker and more Precise in filtering any noise it may be producing. Because the processors demand for current is not constant, having filters as close as possible also is of benefit. This is why we see on package capacitors.

Again this is incomplete and oversimplified compared to the real discussion.
Why don’t power supply reviews include a graph of current along with output voltage ripple? Perhaps even similar readings to the package. And a spectrum analysis.

When power supplies fail, they usually fail spectacularly, and quality should not be underestimated. Higher quality parts make higher quality products. The question you need to ask yourself is: Do you have the ability to troubleshoot every component connected to your power supply?

More germane to the discussion is if noise affects component life. Is seems more data is necessary to reach a conclusion. If its within manufacturer approved tolerances why would it matter? –to component life, not stability of the system-

Why not take a stable overclock and add noise till it fails?
 
It's not hard to test this. Every review site including this one has several cpu's and based on past testing they know exactly how much of an overclock the chip can do on the best motherboards.

Simply test the cpu with powersupplies that are borderline and powersupplies that are known to be stable, and see if the cpu can reach the maximum overclock. The powersupplies this site has plenty of samples of.

What's stopping this site from testing this??
 
Nanotec said:
When power supplies fail, they usually fail spectacularly, and quality should not be underestimated.
Click to expand...
No they don't, unless you mean they stop providing power to your computer is a spectacular failure. Rarely do i ever see a power supply kill a pc and i replace them almost on a daily bases. Power surges and lightning are bigger killers, protecting against these are more important then the power supply itself if you're worried about the life of your system.

People like to overdo it on the power supplies and in my experience. Just a decent power supply will serve you more then well enough for most pc's. Even for overclocking.
 
I will say this... i have gotten the new EVGA supernova 850g2... replacing my OCZ Fatal1ty 750W..which has fried recently... the OCZ was always hott to the touch... this evga supernova is cooler than the other side of the pillow... no hot air coming out of it.... cool electronics means stable and hardy. atleast in my experience...as my always hot OCZ PSU finally failed after about 3 years of service... Put good money into a top of the line PSU..and will power your main rig and your replacement main rigs for a long time.... put the cheapo PSU's in your light work computers.
 
yourgrandma said:
No they don't, unless you mean they stop providing power to your computer is a spectacular failure. Rarely do i ever see a power supply kill a pc and i replace them almost on a daily bases. Power surges and lightning are bigger killers, protecting against these are more important then the power supply itself if you're worried about the life of your system.

People like to overdo it on the power supplies and in my experience. Just a decent power supply will serve you more then well enough for most pc's. Even for overclocking.
Click to expand...

Indeed, even most folks on this site would probably do fine with a high quality 450-500w psu rather than the 800w+ monster they bought.
 
magoo said:
Pretty straight forward.
Quality is better than shit.
Click to expand...

Usually a higher price tag will almost certainly guarantee a better made product when the consumer expect the higher price to perform better compared to less expensive units. The price/performance drop off is considerable at the top 10% (GPU's being the worst with this)

I go with old saying that "You get what you pay for, usually"
 
viscountalpha said:
Usually a higher price tag will almost certainly guarantee a better made product when the consumer expect the higher price to perform better compared to less expensive units. The price/performance drop off is considerable at the top 10% (GPU's being the worst with this)

I go with old saying that "You get what you pay for, usually"
Click to expand...

My friends, whom aren't so knowledgeable with hardware, like to mess with me because I bought a $400~ PSU (Corsair AX1500i) when I would have been fine on a EVGA SuperNOVA 750 G2.

But I had the cash available and I didn't want to ever worry about purchasing a new PSU again, so I bought it. And let me tell you, it feels great knowing that I have as much headroom as I want to play around with overclocking and expanding with more graphics cards.

No more "do I have enough power to add more voltage?". And it also feels good knowing you have high-end quality parts inside.
 
alabrand1 said:
My friends, whom aren't so knowledgeable with hardware, like to mess with me because I bought a $400~ PSU (Corsair AX1500i) when I would have been fine on a EVGA SuperNOVA 750 G2.

But I had the cash available and I didn't want to ever worry about purchasing a new PSU again, so I bought it. And let me tell you, it feels great knowing that I have as much headroom as I want to play around with overclocking and expanding with more graphics cards.

No more "do I have enough power to add more voltage?". And it also feels good knowing you have high-end quality parts inside.
Click to expand...

And that 300 dollar difference could have upgraded you from a 970 to a 980, a 1tb SSD, etc etc etc. An extreme overkill PSU purchase is extremely stupid when the rest of your hardware isn't overkill to match it.
 
You should also have more headroom than you use. I psu converts AC to DC and there is always some loss to heat due to efficiency... that is the reason behind the 80+ certification and so forth. You can cheat that some what with having a psu capable of higher wattage before the components start to wear out or over volt. Basically figure you want 1.2 times what you use so when you add a new gpu or cpu you don't have to buy a new psu. that said most people will never use more then 1K psu and still have plenty of head room. If you run 16GB of system memory, four or more hard drives, a GPU with 3 GB+ of graphic ram, or a more power hungry cpu like a hex core or greater you are up in the 1k or greater... the thing to remember is that a more efficient psu will require less current from the wall so that if you have a nice system or better you may save money in the long run if you own the psu for a couple years. Note electricity is pretty cheap so if you expect to save hundreds of dollars buy less computer lol...

One thing that was not covered is switching psu are better at giving you cleaner power to your components I'm not sure how many psu are still of the non-switching variety but those that are listed as power factor or a switching psu are a lot better when converting dirty or rough current into clean power.
 
drakken said:
You should also have more headroom than you use. I psu converts AC to DC and there is always some loss to heat due to efficiency... that is the reason behind the 80+ certification and so forth. You can cheat that some what with having a psu capable of higher wattage before the components start to wear out or over volt. Basically figure you want 1.2 times what you use so when you add a new gpu or cpu you don't have to buy a new psu. that said most people will never use more then 1K psu and still have plenty of head room. If you run 16GB of system memory, four or more hard drives, a GPU with 3 GB+ of graphic ram, or a more power hungry cpu like a hex core or greater you are up in the 1k or greater... the thing to remember is that a more efficient psu will require less current from the wall so that if you have a nice system or better you may save money in the long run if you own the psu for a couple years. Note electricity is pretty cheap so if you expect to save hundreds of dollars buy less computer lol...

One thing that was not covered is switching psu are better at giving you cleaner power to your components I'm not sure how many psu are still of the non-switching variety but those that are listed as power factor or a switching psu are a lot better when converting dirty or rough current into clean power.
Click to expand...

Lots of misinformation here...

Power supplies are rated on output, not input. 80+ has absolutely nothing to do with it.

Most PSUs have cooling that is temperature-based, not load-based. Meaning, they will usually heat up to about the same temperature before the fan kicks in and keeps them at that temperature. Thus, temperature of internals is not a valid reason to get more efficient PSUs either.

RAM means nearly nothing in terms of power consumption, and thus should not be used in determining PSU sizing. This is especially true of GPU RAM.

Hard drives sip power. Hard drive power consumption should only be considered when hitting 10+ spinners.

Since APFC is a requirement of being 80+ certified, you will not find any switching PSUs. That said, APFC/switching does not make power delivery cleaner or dirtier. That is dependent on the power coming into the PSU.

The only thing you got right was getting approximately 1.2 of what you are estimated to need.
 
Tsumi said:
And that 300 dollar difference could have upgraded you from a 970 to a 980, a 1tb SSD, etc etc etc. An extreme overkill PSU purchase is extremely stupid when the rest of your hardware isn't overkill to match it.
Click to expand...

Depends on how much he plans on upgrading while still on that PSU. He probably could have gotten away with a 1000W PSU for a lot less money and still had room to expand later, but oh well.
 
callous said:
It's not hard to test this. Every review site including this one has several cpu's and based on past testing they know exactly how much of an overclock the chip can do on the best motherboards.

Simply test the cpu with powersupplies that are borderline and powersupplies that are known to be stable, and see if the cpu can reach the maximum overclock. The powersupplies this site has plenty of samples of.

What's stopping this site from testing this??
Click to expand...

What is "stopping" us is that while it may seem simple, it is not. In order to do this, we would need a LOT of resources and time that we don't have. Experimental design for such a thing raises a really nasty set of problems of how to do what is essentially destructive testing without destroying things or tampering with the results by taking measurements with rather large parallel design.

Tsumi said:
That said, APFC/switching does not make power delivery cleaner or dirtier.
Click to expand...

Switching noise ;)
 
Tsumi said:
And that 300 dollar difference could have upgraded you from a 970 to a 980, a 1tb SSD, etc etc etc. An extreme overkill PSU purchase is extremely stupid when the rest of your hardware isn't overkill to match it.
Click to expand...

Way ahead of you, my system is already overkill. There are literally no available upgrade paths for me at the current time.
 
alabrand1 said:
Way ahead of you, my system is already overkill. There are literally no available upgrade paths for me at the current time.
Click to expand...

/facepalm

If your system is way overkill, then the AX1500i would not be overkill, and you wouldn't be having your friends telling you a 750 watt is sufficient. An overkill system would easily pull down 1400+ watts.
 
Paul_Johnson said:
What is "stopping" us is that while it may seem simple, it is not. In order to do this, we would need a LOT of resources and time that we don't have. Experimental design for such a thing raises a really nasty set of problems of how to do what is essentially destructive testing without destroying things or tampering with the results by taking measurements with rather large parallel design.



Switching noise ;)
Click to expand...


I guess we'll never get a straight answer then. Overclocking existed even back in the 90's and the PSU was always the component faulted after the chip, or motherboard yet there was no proof a top of the line PSU made an ounce of difference.

I wish you and kyle could get this special project going as nobody else even tried to tackle this puzzle.
 
callous said:
Overclocking existed even back in the 90's
Click to expand...

Yeah, lot's of us have been around since then and before.

and the PSU was always the component faulted after the chip, or motherboard yet there was no proof a top of the line PSU made an ounce of difference.
Click to expand...

Of course there is if you knew what you were doing or what you were looking at. The qualitative assessment on the margins has always been easy it is the quantitative measure anywhere that is the problem.
 
It is important simply because even if your PSU is slightly degraded, an overclocked machine will create issues. For example, I have a Corsair HX 850. After 4 years, it was working fine until I replaced an "aging" 670 graphics card with a 970. Worked for a couple of weeks until I played a game that really cranked up the fans. I started getting cold boots in the middle of certain games like Rift. Nothing in the event logs. So I thought it was the video card so I RMAed it and the next 970 FTW did the same thing. WTF?

Maybe it's due to my really aging Sandy Bridge so I upgraded to a Z97 build and what do you know, reboots again. So, I have a TX650 I keep around for a standby and after dropping it in, no more reboots. Before I RMAed the HX850, I plugged it in 2 PSU testers - Antec and CoolMax. The Antec tester is all green/red lights and all of them were green. So I plugged in the CoolMax tester which shows the actual readings and I noticed the value on 12v rail at 11.5/11.6. This is with no load too. Hmm. Maybe drop this in a mobo that is highly OCed and it was just enough to trigger it. Keep in mind the 670 never caused any reboots so I never suspected the PSU (since the 970 actually draws slightly less power).
 
callous said:
Overclocking existed even back in the 90's and the PSU was always the component faulted after the chip, or motherboard yet there was no proof a top of the line PSU made an ounce of difference.
Click to expand...


lmfao, man that made me feel old.

But for reference, in the early 2000's I went back and bench oc'd some of my stuff from the 90s, with quality overkill psu's from what was out at the time. Overall my tbirds and athlons definately seem to clock higher and I could POST at speeds I never would have considered back then...

To be fair, the cooling equipment was much better than what I was using in the 90's tho...
 
You must log in or register to reply here.
Back
Top